Fine-grained Semantic Constraint in Image Synthesis

TL;DR

This paper introduces a multi-stage, high-resolution image synthesis model that leverages fine-grained attributes and masks to produce more realistic images with constrained diversity, improving the discriminator and dataset labeling.

Contribution

The paper presents a novel multi-stage model utilizing fine-grained semantic attributes and masks, along with an improved discriminator scheme and dataset labeling optimization.

Findings

Generated images are more realistic and conform to semantic constraints.

The model reduces unexpected diversity in generated samples.

Enhanced discriminator discriminates both entire images and sub-regions.

Abstract

In this paper, we propose a multi-stage and high-resolution model for image synthesis that uses fine-grained attributes and masks as input. With a fine-grained attribute, the proposed model can detailedly constrain the features of the generated image through rich and fine-grained semantic information in the attribute. With mask as prior, the model in this paper is constrained so that the generated images conform to visual senses, which will reduce the unexpected diversity of samples generated from the generative adversarial network. This paper also proposes a scheme to improve the discriminator of the generative adversarial network by simultaneously discriminating the total image and sub-regions of the image. In addition, we propose a method for optimizing the labeled attribute in datasets, which reduces the manual labeling noise. Extensive quantitative results show that our image synthesis model generates more realistic images.